MANAGEMENT SCIENCEVol. 58, No. 2, February 2012, pp. 365–383ISSN 0025-1909 (print) � ISSN 1526-5501 (online) http://dx.doi.org/10.1287/mnsc.1110.1378

© 2012 INFORMS

Do Cultural Differences BetweenContracting Parties Matter?

Evidence from Syndicated Bank LoansMariassunta Giannetti

Stockholm School of Economics, CEPR, and ECGI, SE-113 83 Stockholm, Sweden,mariassunta.giannetti@hhs.se

Yishay YafehSchool of Business Administration, The Hebrew University of Jerusalem, CEPR, and ECGI,

Mount Scopus, 91905 Jerusalem, Israel, yishay.yafeh@huji.ac.il

We investigate whether cultural differences between professional decision makers affect financial contractsin a large data set of international syndicated bank loans. We find that more culturally distant lead banks

offer borrowers smaller loans at a higher interest rate and are more likely to require third-party guarantees.These effects do not disappear following repeated interaction between borrower and lender and are economicallysizable: A one-standard-deviation increase in cultural distance, approximately the distance between Canada andthe United States or between Japan and South Korea, is associated with a 6.5 basis point higher loan spread; theloan spread increases by about 23 basis points if the bank-firm match involves culturally more distant parties,for example, from Japan and the United States. We also find that cultural differences not only affect the relationbetween borrower and lender, but also hamper risk sharing between participant banks and culturally distantlead banks.

Key words : financial contracts; risk sharing; behavioral bias; cultureHistory : Received May 10, 2010; accepted April 16, 2011, by Brad Barber, Teck Ho, and Terrance Odean,

special issue editors. Published online in Articles in Advance July 15, 2011.

1. IntroductionPsychologists and management scholars documentthat national culture affects codes and norms usedduring negotiations and that, consequently, in sim-ulated negotiations between parties with differentnational cultures, joint gains are lower than in nego-tiations between parties that share the same culture(Brett and Okumura 1998, Adair et al. 2001). Evidenceon whether the outcomes of real life negotiations areindeed affected by cultural differences is sparse. Thefrequent failures of mergers involving organizationswith different national cultures would suggest thatthey do (e.g., Weber et al. 1996). An effect of cul-tural differences between contracting parties on con-tract terms would also suggest that common codesand norms (or the lack thereof) play an important rolein actual negotiation outcomes.

This paper examines whether financial contractswritten by parties with different national cultures areaffected by the extent of cultural differences. Nationalcultures may matter for several reasons. First, com-munication is more effective when the source andthe receiver share codes and norms, which is morelikely to happen if individuals share the same culture(Rogers and Bhowmik 1970). Second, national cultureis related to the organizational structure of companies

and affects, for instance, how centralized they are(Bloom et al. 2009). Similar organizations may com-municate and cooperate more easily.

In a large sample of international syndicated bankloans, we show that the bigger are the cultural dif-ferences between the countries of the syndicate’s leadbank and of the borrower, the less favorable arethe loan terms for the borrower. Ceteris paribus,more culturally distant borrowers are offered loansat a higher interest rate, are more likely to need aguarantor, and receive smaller loans. These effectsare economically sizable as a one-standard-deviationincrease in cultural distance, approximately the dis-tance between Canada and the United States orbetween Japan and South Korea, is associated witha 6.5 basis point increase in the loan spread; theloan spread increases by about 23 basis points (or15% of the sample median) if the bank-firm matchinvolves culturally more distant parties, for example,from Japan and the United States. Importantly, theeffects of cultural differences do not disappear if cul-turally distant banks lend repeatedly to a particularborrower or if the lender has a subsidiary in the coun-try of the borrower.

Because negotiations between lead banks and par-ticipant banks parallel negotiations between lead

365

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Giannetti and Yafeh: Do Cultural Differences Between Contracting Parties Matter?366 Management Science 58(2), pp. 365–383, © 2012 INFORMS

banks and borrowers, we also explore the extentto which cultural differences affect the interactionbetween the banks participating in the syndicate.A one-standard-deviation increase in the cultural dis-tance between a participant bank and the lead bankincreases the difference between the lead bank’s por-tion of the loan and the share of the loan heldby the participant bank by 5%, suggesting that cul-tural differences reduce risk sharing within the syndi-cate. Repeated interaction between banks lowers theimpact of cultural differences; however, the negativeeffect of cultural distance on within-syndicate risksharing disappears only after more than 30 joint deals.This is a rare occurrence as 75% of all banks in thesample are involved in 10 joint deals or less.

We thoroughly investigate whether differences infinancial contracts may arise from the fact that cul-turally distant banks attract less creditworthy borrow-ers using selection models, borrower and lender fixedeffects, fixed effects for the borrower’s and the bank’snationalities, and comparisons across different sub-samples, time periods, and regression specifications.All tests consistently indicate that the more conserva-tive terms offered by culturally distant banks are notdriven by the poor quality of the borrowers. More-over, we explore borrowers’ ex post performance andfind no evidence that, after the loan is granted, theperformance of firms borrowing from culturally dis-tant banks is worse than that of other borrowers.

An interpretation of our findings is that culturaldifferences make negotiations more cumbersome andthus increase contracting costs. The effect may be non-pecuniary if interaction with culturally distant bor-rowers increases the lenders’ disutility from writingthe contract.1 The effect may also be pecuniary asmore time and resources may be needed in writ-ing contracts between culturally distant parties. Whatmatters is that contracting costs appear to be relatedto the culture of the parties involved in the negotia-tions, suggesting that behavioral patterns arising fromthe use of different codes and norms should be incor-porated in contract theory.

Another interpretation of our results is that cul-tural dissimilarities increase the cost of informationgathering (or make information gathering less effi-cient). Having less precise information, culturally dis-tant banks consider borrowers riskier than culturally

1 Nonpecuniary costs are equivalent to taste-based discrimination(Becker 1971). In a similar vein, individuals may focus on (irra-tionally) pessimistic scenarios when they deal with culturally dis-similar counterparties. In this respect, our findings are related toa few recent papers showing that ethnic minorities, female bor-rowers, and less attractive individuals pay higher interest rates forreasons unrelated to their risk (Alesina et al. 2008, Ravina 2008).However, an irrational focus on certain scenarios is unlikely forinternational banks.

closer banks do, and therefore offer loans with morerestrictive contract terms. The persistence of the effectof cultural differences despite repeated interactionand across institutional environments and borrowerswith different levels of opaqueness makes it unlikelythat our results are due solely to asymmetric informa-tion. In addition, we find no evidence that the vari-ance of contract terms offered by culturally distantbanks is lower than that of loans offered by domes-tic banks; this, together with considerable empiricalevidence showing that the clients of culturally distantbanks are similar to the clients of other banks, sug-gests that culturally distant banks are as discerning astheir culturally close peers.

The paper is related to several strands of the liter-ature. The link between culture and economic behav-ior has fascinated social scientists ever since MaxWeber. Whereas most of the papers in the literatureexplore the effects of culture on economic outcomes(see Guiso et al. 2006 for a review), we do not inves-tigate the effects of culture per se, but focus on cul-tural differences. In this respect, our paper is closer tothe literature on cultural differences and the flows offoreign direct investment and international mergers(Kogut and Singh 1988, Ahern et al. 2010, Siegel et al.2011). A related strand of literature initiated by Guisoet al. (2009) explores the effects of trust and showsthat trade and investment flows are larger betweencountries that exhibit higher mutual trust. Especiallyrelated to us is Bottazzi et al. (2007), who provide evi-dence that venture capitalists are less likely to fundentrepreneurs in countries whose citizens they trustless and, if they do, the contracts they use are dif-ferent from the contracts used in countries they trustmore. Unlike the literature on mutual trust, we askwhether cultural similarity eases economic interac-tion. The depth of the syndicated loan market allowsus not only to study a much larger set of countries,but also to explore whether the effects of cultural dif-ferences disappear following repeated interaction.

Our paper is also related to the literature onthe home equity bias. Many studies have shownthat lack of familiarity limits investment (Covaland Moskowitz 1999, Huberman 2001). Familiarityis enhanced by cultural similarity (Grinblatt andKeloharju 2001). Our paper contributes to this litera-ture by introducing a new proxy for familiarity andby showing that it enhances financial flows in theform of corporate debt, not only equity. Furthermore,we document that familiarity affects not only quanti-ties, but also the structure of financial contracts.

Finally, our work is related to papers analyzingthe structure of syndicated loan contracts. Esty andMegginson (2003), Qian and Strahan (2007), and Baeand Goyal (2009) explore how creditor protection andlaw enforcement in the borrower’s country shape

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Giannetti and Yafeh: Do Cultural Differences Between Contracting Parties Matter?Management Science 58(2), pp. 365–383, © 2012 INFORMS 367

financial contracts. We contribute to this literature byshowing that cultural distance also matters.

2. Background and Data Sources2.1. The Syndicated Loan Market and

Cultural DifferencesWe study whether cultural distance affects financialcontracts in the syndicated loan market. A syndicatedloan is jointly extended by a group of banks, includ-ing one (or a few) lead banks and many participantbanks. Prior to signing the loan contract, lead banksassess the quality of the borrowers and negotiateterms and conditions. Once the key terms are in place,participant banks are invited to buy a stake of theloan. As a consequence, the issuance of a syndicatedloan is preceded by lengthy negotiations between bor-rowers and lead banks first, and between lead banksand participant banks afterward. Our focus is on theborrower–lender (lead bank) relation; in §5, we brieflyanalyze the risk sharing between lead banks and par-ticipant banks.

The syndicated loan market is an appropriate con-text to explore the effects of cultural differences onfinancial contracts for several reasons. First, as Duffieet al. (2005 and 2007) highlight, obtaining a syn-dicated bank loan presents search and bargainingfrictions. Because of the length of the negotiationsinvolved, borrowers cannot approach multiple poten-tial lenders contemporaneously and cannot comparemultiple offers. Because of an increasing cost of fundsor capacity constraints, banks are not always availableto extend loans. Borrowers, having opportunity costsof waiting, are therefore not necessarily matched tothe most suitable counterparties.

Second, negotiations are important to establish con-tract terms. As psychologists and management schol-ars point out (Brett and Okumura 1998, Adair et al.2001), sharing similar norms and codes facilitatescommunication and the exploration of alternatives(for instance, a borrower may need long maturity, butmay be willing to concede on the loan amount).

Negotiations may also be made less effective by thefact that, in different cultures, organizations are repre-sented by individuals with different skills and roles.For instance, in Anglo-Saxon individualistic cultures,companies tend to select the most energetic membersfor the negotiations; Chinese or Japanese teams areoften led by a senior person, who has a high statusin the organization and may lose face when dealingwith the younger representatives of the counterparty.Also, negotiators from individualistic and egalitariancultures have the power to accept and reject offers,whereas in more hierarchical cultures the members ofthe organization with actual decision power are notpresent at the meetings. Cultural values also affect

corporate policies such as gender equality, diversity,and (attitudes toward) environmental policies. Sim-ilar ethnic and gender stereotypes or expectationson environmental standards affect how comfortablenegotiators are with the counterparties. All these fac-tors can make cross-cultural negotiations lengthy andineffective and increase contracting costs. After thebank obtains a fair remuneration on its investment,these contracting costs are manifest in less favorableloan terms for the borrower.2

It is also possible that cultural dissimilaritiesincrease the cost of information gathering and that,as a consequence, an identical borrower may be con-sidered riskier by a culturally distant bank. Thismechanism would also suggest that cultural similar-ity facilitates negotiation by enhancing informationsharing. Whereas information asymmetry is known toplay a role in the syndicated loan market (Dennis andMullineaux 2000), the results we present below lendstronger support to a story based on contracting costs.

2.2. Syndicated Loan DataData on syndicated loans are from Dealogic’s Loan-ware Database, which provides information on bor-rowers, lenders, and loan price and nonprice terms atorigination. This database is widely used for studyingthe international syndicated loan market (Esty andMegginson 2003, Carey and Nini 2007).

Whereas Loanware contains information on syn-dicated loans to local and central governments, wefocus on corporate borrowers. We extract informationon contracts from 1980 to 2005. Less than 15% of thecontracts are signed in the first 10 years, reflecting thefact that the syndicated loan market was still under-developed during the 1980s. It is also possible thatLoanware coverage is less complete at the beginningof the period or for some countries. Therefore, in theempirical analysis we make sure that our results donot hinge upon the inclusion of the 1980s or of coun-tries with fewer than 100 loans.

2.3. Measuring Cultural DistanceThe definition of culture usually includes some notionof shared values, beliefs, codes, and norms. The WorldValues Survey (WVS) is an attempt by social scien-tists to measure cultural values around the world. TheWVS initially covered only 22 countries and was con-ducted at ten-year intervals; currently, the survey cov-ers about 80 countries and is updated every five years.The survey consists of a detailed questionnaire on

2 Contracting costs may be pecuniary or nonpecuniary. The lineof demarcation between the two is tenuous. For instance, Becker(1971) recognizes that taste-based discrimination may arise notfrom the prejudice of the employer (the firm), but from the tastesof coworkers who demand to be compensated by higher wages forworking with minorities.

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Giannetti and Yafeh: Do Cultural Differences Between Contracting Parties Matter?368 Management Science 58(2), pp. 365–383, © 2012 INFORMS

Figure 1 Cultural Map of the World

Ireland

U.S.A

N.IrelandAustralia

Canada

New zealand

GreatBritain

IcelandAustria

Switzerland

NetherlandsFinland

WestGermany

Denmark

Sweden

Japan

ChinaEstoniaBulgaria

Russia

Ukraine Belarus

Albania SerbiaSlovakia

Slovenia

Czech

EastGermany

Greece

Israel

CroatiaBosnia

Poland

SouthAsia

Catholic Europe

India

IranBangladesh

Pakistan

Tanzania Colombia

El Salvador

PuertoRico

VenezuelaMexicoGhana

NigeriaEgyptAlgeriaUgandaJordan

ZimbabweMorocco

Africa

Latin AmericaSouthAfrica

Turkey

IndonesiaPhilippines

PeruBrazil

ChilePortugal

ArgentinaDominicanRepublic

Vietnam

Uruguay

Spain

Italy

Belgium

France Luxembourg

LatviaMontenegro

Lith-uania

Moldova

Romania

ArmeniaAxerbajan

Georgia

HungaryMacedonia

S.Korea

Taiwan

Norway

ProtestantEuropeConfucian

Ex-Communist

Englishspeaking

–2.0

–2.0

–1.5

–1.0

–0.5

0

0.5

1.0

1.5

2.0

–1.5 –1.0

Survival values

Trad

ition

al v

alue

s

Factor score

–0.5 0 0.5 1.0 1.5 2.0

Sec

ular

/rat

iona

l val

ues

Self-expression values

Source. World Values Survey, http://www.worldvaluessurvey.org; Inglehart and Welzel (2005, p. 63).

concrete aspects of life (about 250 questions) admin-istered in face-to-face interviews; the average numberof respondents is 1,400 per country.

Inglehart (1997) and Inglehart and Baker (2000)show that diverse orientations tend to cluster togetherin coherent patterns. Consequently, they use factoranalysis to summarize the salient features of differentcultures along two dimensions (values): (1) the extentto which a society emphasizes traditional as opposedto secular values; (2) the extent to which a societyemphasizes values related to survival as opposed toself-expression. In societies with traditional values,individuals emphasize religion, family ties, and def-erence to authority. Survival values are considered tobe predominant in societies with low interpersonaltrust, which tend to be intolerant of ethnic and cul-tural minorities, do not support gender equality orenvironmental protection, and often favor authoritar-ian governments. Besides affecting corporate policieson gender, diversity, environment etc., cultural val-ues are related also to the degree of centralization of

organizations, the identities of the individuals con-ducting the negotiations, and whether they actuallyhave the power to make decisions.

Cultural distance between any pair of countries canbe measured as the Euclidean distance between thetraditional versus secular/rational and the survivalversus self-expression orientations. The cross-countrycultural differences that emerge are summarized in acultural map of the world, reproduced in Figure 1 onthe basis of a recent edition of the survey. Althoughthe time-series variation in cultural distance is lim-ited, whenever possible, we measure culture in theyears that immediately precede the signing date ofthe loan.

We attribute to each borrower the culture of its owncountry and to the lead bank the culture of the coun-try where its headquarters are located for two rea-sons. First, the individuals writing the contracts orthe executives with high decision power are likely tobe nationals of the bank’s and the borrower’s coun-try. Second, the culture of the headquarters’ country

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Giannetti and Yafeh: Do Cultural Differences Between Contracting Parties Matter?Management Science 58(2), pp. 365–383, © 2012 INFORMS 369

affects organizational culture and the degree of cen-tralization of the subsidiaries (Bloom et al. 2009).

Because religion has an important role in shapingcultural values, we also use a dummy variable thattakes the value one if the countries of the borrowerand of the lead bank share the same religion as aproxy for cultural similarity. In Giannetti and Yafeh(2010), we also show that our results are robust tothe use of alternative measures of cultural distance,constructed by Hofstede (2001) and Schwartz (2006).In particular, Hofstede’s “power-distance” score cap-tures the centralization of decision power and allowsus to measure cultural differences related to orga-nizational structure. Finally, in Giannetti and Yafeh(2010), we also examine to what extent cultural dis-tance captures trust between nations. We run a horserace between our measure of cultural distance and theproxy for trust proposed by Guiso et al. (2009) andfind that, although our proxy for cultural distance ispositive and statistically significant, trust is not, indi-cating that cultural differences matter beyond trust.3

3. Main Variables and DescriptiveStatistics

Our sample includes about 86,000 loans to over 40,000borrowers in more than 70 countries from 1980 to2005. There are more than 6,500 lead banks fromnearly 60 countries and over 8,000 participant banks.The list of the largest borrower and lead bank nation-alities and the cultural distance between them is pre-sented in panel A of Table 1.

Panel B of Table 1 describes ex ante loan character-istics. In most of the analysis, we focus on the loanspread, which is measured as the basis point spreadover the LIBOR (London Interbank Offered Rate),inclusive of all fees. Our results, however, are robust ifwe use a measure of the cost of the loans that excludesfixed fees. Nonprice terms also affect the lender’s abil-ity to obtain a fair return on the loan; these includethe loan amount, its maturity, and whether the loanis secured or guaranteed by a third party. Finally, wealso observe the borrower’s credit rating (if any) at thetime the loan is granted and any subsequent changesin rating before the maturity of the loan, includingwhether the loan is downgraded to default.

Loanware also provides information on the identityof the lead banks and their nationalities, as well ason the composition of the syndicate. For over 75% of

3 We also explore whether the effect of cultural distance is asym-metric. First, we consider whether cultural distance matters moreif the borrower is in a weaker creditor protection country than thelender. Second, we consider whether borrowers from countries thattend to stress more traditional and survival values obtain worseloan terms. In all these unreported tests, we find that the effect ofcultural distance is symmetric.

the loans in our sample, there is only one lead bank.We thus consider the lead bank as the lending bank(as is customary in the literature) and use the leadbank’s nationality to define cultural distance from theborrower and all the other lead bank’s nationality-based variables. In the few cases in which there areseveral lead banks, to be as conservative as possible,we define all the variables with respect to the leadbank that is culturally closest to the borrower.4

Panel C of Table 1 presents the various measuresof distance between borrower and lender. Beside themeasure of cultural distance from the WVS, we alsouse physical distance and the absolute value of thedifference between the index of creditor rights in theborrower’s and the lead bank’s countries. Moreover,in most of regressions, we include four dummy vari-ables (not tabulated) capturing, respectively, coun-tries with a common border, with a common legaltradition, with the same language, and with com-mon colonial ties. These features have been shown tofavor international trade (Rose 2004). Admittedly, lan-guage and common history capture aspects of culture.Nevertheless, we include these variables as controlsbecause the aspects we want to capture are not nec-essarily related to the spoken language or to colonialhistory.5

Panel D of Table 1 summarizes the salient featuresof the syndicate composition. Our goal here is toexplore the extent to which risk sharing within thesyndicate depends on the cultural distance betweenthe lead bank and each of the participants. In a situ-ation of perfect risk sharing, all (similar) banks in thesyndicate would equally fund the loan. However, ifnegotiations between culturally distant banks are lesseffective, the extent of risk sharing may be lower. Wedefine actual risk sharing in the syndicate as the loanamount provided by a given participant bank stan-dardized by the loan that each bank in the syndicatewould provide under perfect risk sharing, minus theloan amount provided by the lead bank, also stan-dardized by the loan that each bank would extendunder perfect risk sharing. This variable does notdepend on the total size of the loan and on the num-ber of syndicate participants and therefore allows usto measure a participant bank’s willingness to sharerisk with a particular lead bank. As in the case of leadbank–borrower relations, we define distance variables

4 Our results are unchanged if we restrict the sample to syndicatedloans with one lead bank only.5 For instance, management scholars suggest that negotiators fromFrance or Belgium should expect greater problems in cooperatingwith negotiators from Denmark, New Zealand, or the United King-dom than with negotiators from, say, Korea or El Salvador becausethe culture of the latter stresses authority to a similar large extent(Mead and Andrews 2009).

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Giannetti and Yafeh: Do Cultural Differences Between Contracting Parties Matter?370 Management Science 58(2), pp. 365–383, © 2012 INFORMS

Tabl

e1

Varia

ble

Defin

ition

san

dSa

mpl

eSt

atis

tics

Borr

ower

s

Lend

ers

AUBR

ZCA

CHN

FRGR

HKIN

DIA

INDO

ITJP

NM

XNL

NOR

SING

KOR

SPSW

ETU

RUK

USTo

tall

ende

rs

Pane

lA.N

umbe

rofl

oans

tom

ajor

borr

owin

gco

untri

esAu

stra

lia(A

U)63

510

1518

101

875

(0.2

)(2

.6)

(0.7

)(0

.6)

Braz

il(B

RZ)

158

017

6

Cana

da(C

A)46

111,

574

1597

1,47

03,

306

(0.2

)(2

.0)

(1.0

)(0

.4)

(0.4

)

Chin

a(C

HN)

189

293

1321

540

(0)

(1.5

)(0

.21)

Fran

ce(F

R)13

2027

3961

041

2976

3930

1716

7648

5498

671

2,09

7(1

.3)

(1.8

)(0

.9)

(2.1

)(1

.7)

(1.7

)(2

.2)

(0.5

)(1

.4)

(1.7

)(0

.8)

(1.2

)(0

.9)

(1.8

)(0

.5)

(1.4

)

Germ

any

(GR)

3116

2470

1977

772

2450

3745

1314

108

2211

237

2136

143

1,17

73,

388

(1.7

)(2

.6)

(1.6

)(1

.8)

(0.8

)(1

.7)

(2.2

)(2

.5)

(1.0

)(0

.8)

(2.6

)(1

.2)

(0.3

)(2

.0)

(1.3

)(1

.2)

(0.9

)(2

.4)

(1.2

)(2

.1)

Hong

Kong

(HK)

4486

140

(0)

Indi

a(IN

DIA)

213

242

Indo

nesi

a(IN

DO)

208

220

Italy

(IT)

1523

1513

1,28

811

4154

1663

1,74

4(1

.7)

(1.9

)(1

.6)

(1.8

)(1

.2)

(0.4

)(1

.5)

(0.5

)(1

.0)

Japa

n(J

PN)

107

2520

167

2938

268

5315

597

256

1416

4141

355

2823

116

1,00

13,

150

(2.4

)(3

.2)

(2.2

)(1

.6)

(1.4

)(0

.9)

(1.5

)(2

.5)

(3.1

)(1

.6)

(1.9

)(1

.0)

(2.4

)(1

.0)

(1.8

)(1

.6)

(2.8

)(1

.9)

(2.8

)

Mex

ico

(MX)

8085

Neth

erla

nds

(NL)

1015

1912

3442

1450

518

1017

1134

986

1,88

4(1

.0)

(2.9

)(1

.0)

(2.7

)(2

.7)

(3.1

)(2

.7)

(0.6

)(2

.6)

(2.5

)(0

.8)

(0.9

)(1

.4)

Norw

ay(N

OR)

267

4934

2(2

.0)

Sing

apor

e(S

ING)

5338

127

222

552

(1.5

)(0

.3)

(0.7

)

Sout

hKo

rea

(KOR

)61

7369

789

451,

098

(0.5

)(0

.2)

(2.2

)(2

.7)

Spai

n(S

P)34

1866

1,16

025

1922

1,61

7(1

.4)

(0.4

)(1

.1)

(1.0

)(0

.8)

(1.3

)

Swed

en(S

WE)

402

538

Turk

ey(T

UR)

129

142

Unite

dKi

ngdo

m(U

K)50

2155

3328

2818

529

6225

1320

3039

8163

134,

093

848

5,91

3(0

.6)

(2.0

)(0

.4)

(2.5

)(0

.5)

(1.1

)(2

.2)

(2.0

)(2

.2)

(0.6

)(1

.8)

(1.6

)(0

.9)

(1.8

)(2

.1)

(0.9

)(1

.9)

(0.9

)

Unite

dSt

ates

(US)

102

115

229

1726

2656

4711

932

3826

535

2752

6090

238

52,6

5054

,909

(0.6

)(1

.7)

(0.4

)(3

.1)

(1.3

)(1

.9)

(2.8

)(2

.2)

(2.1

)(1

.2)

(2.6

)(1

.3)

(1.4

)(1

.9)

(2.6

)(1

.3)

(1.7

)(0

.9)

Tota

lbor

row

ers

1,02

950

32,

048

719

755

906

1,19

851

697

41,

615

421

525

679

457

427

1,62

61,

540

504

621

5,08

559

,810

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Giannetti and Yafeh: Do Cultural Differences Between Contracting Parties Matter?Management Science 58(2), pp. 365–383, © 2012 INFORMS 371

Table 1 (Continued)

Variable Definition/source Units Mean Std. dev. 25% Median 75% Obs.

Panel B. Contract characteristicsSpread Loan cost including all fees/Loanware Basis points p/a 189 219 6205 150 250 861354

above LIBOR

Amount Loanware Million USD 19005 51702 20 60 175 1161803Maturity Loanware Years 403 302 2 4 505 1011202Secured Dummy that takes the value 1 if the loan is

secured/Loanware0/1 0029 0045 0 0 1 1171194

Guaranteed Dummy that takes the value 1 if the loan isguaranteed/Loanware

0/1 0007 0026 0 0 0 1171194

Tranched Dummy that takes the value 1 if the loan is offeredin several separate tranches/Loanware

0/1 0046 0049 0 0 1 861354

Number of banks Number of banks in the syndicate 706 901 2 4 10 861354

Foreign bank Dummy that takes the value 1 if the firm borrowsfrom a foreign bank/Loanware

0/1 0022 0041 0 0 0 861354

Borrower interaction Total number of loans (including current) of the leadbank to the borrower/Loanware

105 101 1 1 2 861354

Rating Borrower credit rating at the time the contract issigned on a scale from 1 (AAA) to 21 (C orlower); refers to the lower of Moody’s and S&P’sratings, if both are available

1 to 21 1009 404 7 11 15 251202

Default Proportion of borrowers classified by Moody’s orS&P’s as in default

Percentage 005 007 0 0 0 861354

Panel C. Measures of distance between borrower and lead bankCultural distance Euclidean distance between the cultures of the

borrower’s and the lead bank’s countries/WVSSee text for details 0030 0068 0 0 0 861354

Distance Physical distance between the capital of the countryof the lead bank’s headquarters and the capital ofthe borrower’s country/Infoplease.com

1,000 km 1021 3027 0 0 0 861354

Creditor rights distance Absolute value of the difference between creditorrights in the lead bank’s country and in theborrower’s country/Djankov et al. (2007)

0 to 4 0027 0070 0 0 0 861354

Panel D. Syndicate composition and characteristicsRisk sharing (Loan held by participant i)/(loan amount/number of

banks) − (loan held by the lead bank)/(loanamount/number of banks)/Loanware

−203 17092 −1082 −0098 −005 2251704

Interaction syndicate Number of previous deals of a participant bank witha lead bank, including current/Loanware

7099 11034 1 2 9 2251704

Banks’ cultural distance Cultural distance between the participant bank’s andlead bank’s countries/WVS

See text for details 0068 0088 0 0022 1017 2251704

Banks’ distance Physical distance between the capital of the countryof the lead bank’s headquarters and the capital ofthe country of the participant bank’sheadquarters/Infoplease.com

1,000 km 2079 4018 0 003 5086 2251601

Creditor rights distancesyndicate

Absolute value of the difference between creditorrights in the participant bank’s country and in thelead bank’s country/Djankov et al. (2007)

0 to 4 0064 0093 0 0 1 2251704

Notes. Panel A presents the top 20 borrowers and lenders’ nationalities in the sample. The nationality of the lead lender is listed in the rows and borrower nationalityis listed in the columns. The total figures in the rows (columns) include all loans from lenders (to borrowers) in each country, not only the ones involving the top 20borrowing nations. Cultural distance, from the World Values Survey, appears in parentheses. The figures are time varying and are calculated for the years in whichcontracts are signed; therefore, the average cultural distance between, for example, lenders from France and borrowers from Germany, need not be exactly equal to thecultural distance between lenders in Germany and borrowers in France. The other countries included in the sample, but not reported in panel A either as lenders or asborrowers are Algeria, Argentina, Austria, Bangladesh, Belarus, Belgium, Bulgaria, Chile, Colombia, Croatia, Czech Republic, Denmark, Egypt, El Salvador, Finland, Ghana,Greece, Hungary, Iran, Ireland, Israel, Jordan, Latvia, Lithuania, Morocco, New Zealand, Nigeria, Pakistan, Peru, Philippines, Poland, Portugal, Puerto Rico, Romania,Russia, Saudi Arabia, Slovakia, Slovenia, South Africa, Switzerland, Tanzania, Ukraine, Uruguay, Venezuela, Vietnam, and Zimbabwe. The statistics in all panels of Table 1are calculated using only observations included in subsequent tables. For variables used in the analysis of loan spreads, we use only observations included in column(5) of Table 3, which we use as a benchmark in most of the subsequent tests. For other variables, we use only observations included in the regression where the variableis used.

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Giannetti and Yafeh: Do Cultural Differences Between Contracting Parties Matter?372 Management Science 58(2), pp. 365–383, © 2012 INFORMS

using the countries where the headquarters of the par-ticipants and of the lead bank are located. As above, ifthere are multiple lead banks, we select the lead bankthat is culturally closest to a given participant.

Besides the loan ratings, the following untabulatedloan characteristics allow us to control for borrowerheterogeneity: 56 industry dummies; 21 dummiescapturing the loan purpose (e.g., whether the loan isneeded to finance an acquisition, to buy a specificasset, or as working capital); and 11 borrower typedummies capturing whether the borrower is publiclyor privately owned and whether it is a bank, anothertype of financial institution, a utility company, ora company in another industry. All these borrowercharacteristics and, in particular, the credit rating, cap-ture differences in the risk of firm assets and capitalstructure (Kisgen 2006). We also include 46 dummiescapturing the loan instrument type (e.g., whether theloan is a credit line, a term loan, a bridge facility etc.)and 69 currency dummies.6

In addition, we match by name Loanware firmswith Worldscope to obtain financial statements fora subsample of large listed borrowers. For this sub-sample, whose size is comparable to the sample ofQian and Strahan (2007), we have information onsales, percentage of foreign sales, the market to bookratio, profitability (net income over assets), and theproportion of tangible assets (property, plants andequipment) over total assets. Finally, we collect time-varying country controls, including a proxy for thesupply of credit in the borrower’s country (credit togross domestic product (GDP)) as well as creditorrights and GDP per capita in the countries of both theborrower and the lead bank.

A major question is whether domestic banks, cul-turally close banks, and culturally distant (foreign)banks attract different types of borrowers. In Table 2,we define culturally close banks as those whose cul-tural distance is below the mean of the subsampleof foreign banks and show how contract terms andborrower characteristics vary across lending banks.Loans extended by culturally distant banks have,on average, lower spreads than loans extended bydomestic banks, but are more expensive than the onesextended by culturally close (foreign) banks. Becausewe include no controls, this evidence is not neces-sarily inconsistent with our conjecture. Even thoughon average foreign banks are able to extend cheaperloans than domestic banks, they may provide morerestrictive loan terms to culturally distant borrow-ers than to the average of their other clients. Cultur-ally distant banks appear to be more likely to extendsmaller loans and to require guarantees or collateral.

6 Even though some of these are endogenously chosen contract fea-tures, their inclusion may help capture the risk of the loan. Theomission of subsets of the dummies does not affect our estimates.

Differences in borrower characteristics across dif-ferent subsamples are economically quite small. Forinstance, the average rating of loans issued by domes-tic banks is slightly lower than the average rating ofloans issued by culturally distant and especially cul-turally close foreign banks. To the extent that anyunobserved heterogeneity is correlated with the ini-tial rating of the loans, having domestic banks, forwhich cultural distance is zero, as the bulk of the sam-ple makes it more difficult to find a negative effectof cultural distance on contract terms. Similar conclu-sions on the direction of any unobserved heterogene-ity can be drawn from the ex post performance ofthe loans issued by different groups of banks. Focus-ing on rated borrowers and on loans whose ratingchanges, the proportion of upgraded loans is largerfor loans issued by culturally distant banks than forloans issued by domestic banks. Also, culturally dis-tant banks do not appear to lend to firms with morevolatile performance as the rating of a smaller frac-tion of their borrowers is changed after the grantingof the loan.

Finally, firms borrowing from foreign banks havesimilar size and profitability. Most importantly, incomparison with firms borrowing from domesticlenders, firms borrowing from culturally distantbanks have lower leverage ratios and more tangibleassets, suggesting that they are more creditworthy.Overall, Table 2 provides no indication that culturallydistant banks systematically attract worse borrowersand makes it unlikely that unobserved heterogeneitydrives our results.

4. The Effects of Borrower–LenderCultural Differences onLoan Contracts

4.1. Empirical ApproachWe estimate reduced form equations. Besides culturaldistance and the controls described in §3, in all equa-tions, we include dummies for the borrower’s andlead bank’s nationalities, which may systematicallyaffect contract terms. For instance, the expected repay-ment may be systematically lower for borrowers incountries with weak creditor protection (Qian andStrahan 2007). Similarly, the cost of extending a loanmay be systematically higher for banks from coun-tries with higher funding costs. We also include yeardummies to control for differences in credit marketconditions over time.7

7 In unreported specifications, we include interaction terms of bor-rower nationality and year dummies thus controlling for any pos-sible changes in the borrower’s economic environment. The effectof cultural distance is similar to the one we report.

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Giannetti and Yafeh: Do Cultural Differences Between Contracting Parties Matter?Management Science 58(2), pp. 365–383, © 2012 INFORMS 373

Table 2 Characteristics of Firms Borrowing from Domestic, Culturally Close, and Culturally Distant Banks

Domestic Low cultural High culturalbanks distance bank distance banks

Observations 68,084 8,841 9,429

OutcomesSpread 193∗ 169∗ 180

(150) (108) (131)[226] [195] [190]

Loan amount 217∗ 159∗ 125(73) (60) (50)

[570] [400] [387]Loan maturity 4.2∗ 4.8∗ 4.6

(4.0) (5.0) (4.0)[2.9] [3.5] [3.5]

Secured (%) 31∗ 26∗ 35Guaranteed (%) 5∗ 12∗ 18

Ratings and rating changesAverage rating (on a scale from 1 (AAA) to 21 (C)) 11.1∗ 9.7∗ 10.4

(11) (9) (10)[4.4] [4.4] [4.6]

Unrated (%) 70.4∗ 71.3∗ 75.8Proportion of upgraded firms out of 48.2∗ 50.4% 51.5%

all firms for which the ratingchanged after the granting of the loan

Proportion of firms whose rating changed 29.6∗ 27.9∗ 25.2after the granting of the loan (%)relative to firms with unchanged rating

Proportion of unrated firms that obtained 10.7 11.2 10.6a rating after the granting of the loan (%)

Default (%) 0.4 0.7 0.5

Firm attributesObservations 10,371 1,353 1,171Sales (million U.S. dollars) 3,175,593 3,871,536 3,029,002

(645,878) (1,186,925) (1,511,165)[10,500,000] [12,600,000] [4,393,677]

Leverage 0.34∗ 0.34∗ 0.32(0.30) (0.31) (0.30)[0.38] [0.26] [0.21]

Net income over assets 0.08 0.09 0.08(0.11) (0.10) (0.11)[0.42] [0.29] [0.32]

Property, plant, and equipment 0.37∗ 0.38∗ 0.42(PP&E) over assets (0.31) (0.36) (0.44)

[0.27] [0.31] [0.30]

Notes. This table presents sample means, medians (in parentheses), and standard deviations (in square brackets). Lowcultural distance is defined as positive cultural distance below the sample mean that is about 1.4 for the subsample offoreign banks and high cultural distance is defined as cultural distance exceeding 1.4. We present statistics for contractterms, initial ratings, and rating changes (upgrades or downgrades after the loan was granted) and firm characteristics.

∗Denotes differences in means statistically significant at the 5% level relative to the means reported in column (3).

Our extensive set of controls should capture bor-rower heterogeneity. Thus, any effect of cultural dis-tance on loan terms should be interpreted as arisingfrom culturally distant banks’ policies toward (simi-lar) borrowers. However, it is important to stress thatthe ordinary least squares estimates of the effect ofcultural distance on loan terms rest on the assump-tion that bank and borrower characteristics unrelatedto cultural distance drive the matching of borrowersand lenders. As discussed in §2.1., this identifying

assumption is consistent with the organization ofthe syndicated loan market, in which borrowers facesearch and bargaining frictions. To the extent thatthe nonrandom selection of borrowers could affectour estimates, the direction of the bias may beagainst finding any negative effect of cultural dis-tance on contract terms as, for instance, in Table 2we find that the clients of culturally distant banksare more likely to be upgraded after the loan isgranted.

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Giannetti and Yafeh: Do Cultural Differences Between Contracting Parties Matter?374 Management Science 58(2), pp. 365–383, © 2012 INFORMS

Nevertheless, we explicitly examine whether unob-served heterogeneity biases our estimates usingtwo alternative methodologies. First, we includeborrower-fixed effects and explore whether culturaldistance affects the terms at which the same bor-rower obtains loans from different banks. Second, weconsider a two-stage selection model. This involvesmodeling the probability that a firm obtains the loanfrom a given lead bank and including the inverseMills ratio obtained from the estimated probabilityin the second-stage equation. In the first stage, wehypothesize that a given borrower could obtain a loanfrom any of the domestic and foreign lead banks thatever extended a syndicated loan to borrowers in thesame country up to the year in which the contractis signed. We estimate the probability of observing amatch between a particular lead bank and a particu-lar borrower as a function of borrower, country, andlead bank characteristics.8

To capture the variation in the probability of abank-firm match that is independent of borrowercharacteristics and directly test for the existenceof a matching mechanism driven by search andbargaining frictions, we posit that the probability of abank-firm match depends on the distribution and thenumbers of banks active in a country. In particular,we include the bank’s rank in the country (obtainedby counting the number of deals the bank completedup to the year in which the contract is signed) andthe number of physically and culturally close banks(banks with physical or cultural distance below themean). Similarly to Sorensen (2007) and Bottazzi et al.(2008), our identifying assumption is that the charac-teristics of the other investors in the market shouldnot be directly related to the cost of funding, aftercontrolling for the aggregate supply of credit.

4.2. Loan SpreadTable 3 shows that the effect of cultural distance onthe loan spread is consistently positive and signifi-cant. Because we include lead bank nationality dum-mies, a positive effect of cultural distance on the costof the loan does not necessarily imply that the bor-rower receives funding at a higher absolute cost froma culturally distant bank than from a domestic one.For example, French banks on average extend loansat a lower interest rate than domestic banks to cul-turally distant U.S. borrowers and, at the same time,offer worse contract terms to U.S. borrowers thanto culturally closer Belgian borrowers. The effect of

8 In these tests, to keep the size of the data set manageable, we ranklead banks according to the loans issued up to the contract yearin each country and keep in the sample at most the top 500 activelead banks; any loans extended by lead banks that are not amongthe top 500 are excluded. Different cutoffs (50, 100, 200, and 300)yield similar results.

cultural distance is also economically significant. Incolumn (5), the benchmark specification with a com-prehensive set of controls, a one-standard-deviationincrease in cultural distance, approximately the differ-ence between Canada and the United States, increasesthe spread by approximately 6.5 basis points, or about4% of the sample median spread of 150 basis points.

We explore whether borrower unobserved hetero-geneity leads us to overestimate the effect of culturaldistance by including different sets of controls for bor-rower heterogeneity. To the extent that unobservedheterogeneity in borrower characteristics is correlatedwith the observed controls, the coefficient estimatesshould vary a lot across columns. In fact, the coeffi-cient estimates are very similar, when we include nocontrols for borrower rating (column (1)), when wecontrol for rating by including 4 or 14 rating groups(columns (2) and (3), respectively). In column (4),we add controls for a number of loan characteristics.The latter are admittedly jointly determined with theinterest rate; yet they help in further controlling forborrower heterogeneity, for loan size, and for the pos-sible effects of risk sharing within the syndicate.9 Thecoefficient of cultural distance remains unaffected.

In column (5), we include other controls for dis-tance. In line with our maintained hypothesis, thespread is slightly lower if we include the same reli-gion dummy, an alternative proxy for cultural simi-larity. Sharing the same language or colonial historyseems largely irrelevant. Other aspects of remotenesssuch as the physical distance between the capital citiesof the borrower’s and the lead bank’s countries donot have a significant effect. This is probably becausemany lead banks have subsidiaries in the countryof the borrower or in nearby countries, which maymitigate the effect of geographical, but not of cul-tural distance; we revisit this issue in §4.5. Interest-ingly, differences in creditor rights between the coun-tries of the borrower and of the lender increase loanspreads. However, the effect has only weak statisticalsignificance.

In column (6), we include only loans extended byforeign lead banks; our estimates are qualitativelyunchanged, showing that the results are not driven bythe difference between domestic and foreign banks.The results are also unchanged if we include lenderfixed effects (column (7)). In addition, the effect of cul-tural distance remains unchanged when lead banksfrom the United States or the United Kingdom areexcluded (results not reported) suggesting that theeffect is not driven by the behavior, or market power,of the largest and most reputable banks, which tend to

9 The nationalities of banks participating in the syndicate withoutleading it are not expected to affect the loan terms, which are deter-mined by the lead bank before other participants join the syndicate.

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Giannetti and Yafeh: Do Cultural Differences Between Contracting Parties Matter?Management Science 58(2), pp. 365–383, © 2012 INFORMS 375

Table 3 The Determinants of Loan Spreads

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)

4 rating 14 rating Contract Foreign Lender FirmNo rating groups groups terms Other banks fixed No U.S. level Ratedcontrols controls controls controls distances only effects borrowers controls borrowers

Cultural distance 12086∗∗∗ 13009∗∗∗ 13011∗∗∗ 11015∗∗∗ 80692∗∗ 11020∗∗∗ 11057∗∗∗ 17044∗∗∗ 12099∗ 23081∗∗∗

4301725 4301905 4301875 4205335 4401305 4401695 4401965 4404735 4706125 4406795Distance −000339 −00974 00263 −000193

4006825 4007545 4007175 4007735Common border −10004 −23004∗∗∗ −30184 −70303

4609495 4708935 4803715 4704145Same legal origin 70607 80573∗ 50716 60440

4409125 4408875 4600495 4406865Same religion −70806∗ −10081∗∗ −90325∗ −90781∗

4403915 4408235 4502845 4502465Same language −50431 −0000283 −10895 −70230

4608575 4701635 4703495 4607095Common colonial ties 30699 −12030 20633 12072

4601735 4170125 4704255 4800005Creditor rights—borrower −10056 −10087 −12042 −50233

4705905 4808485 4800905 4707775Creditor rights—lead bank 16077∗∗∗ 80987 12087∗ 80041

4601635 4603795 4704045 4703405Creditor rights distance 20293 30182 40905∗ 40512

4207925 4208515 4207235 4303255Creditor rights are better in lender country dummy −80358 −80594 −18044∗∗ −90608

4908675 4902785 4900125 4100865Credit to GDP—borrower 000498 000535 000407 000424 000610 00320∗∗∗ 000283 00106 000646 −00132

40008735 40008865 40008845 40006855 40006685 4001165 40007415 40007685 4001415 4001125Per capita GDP—lead bank −4052∗∗∗ −4048∗∗∗ −40359∗∗∗ −40006∗∗∗ −4099∗∗∗ −4067∗∗∗ −30199∗∗ −3070∗∗∗ −13041∗∗ −7049∗∗∗

4107425 4107105 4106435 4103605 4104145 4107905 4103965 4103475 4502375 4203675Per capita GDP—borrower 10747 10070 10165 00449 10203 −00915 −00808 −10423 −10787 60404∗∗

4107895 4107135 4106605 4103035 4103605 4104735 4102655 4103885 4502275 4209095Tranched 25061∗∗∗ 25006∗∗∗ 23016∗∗∗ 22050∗∗∗ 25027∗∗∗ 16050∗∗∗ 25000∗∗∗ 12008∗∗∗ 11068∗∗∗ 25017∗∗∗

4206445 4206385 4202495 4405815 4409975 4306135 4106315 4205255 4400255 4106585Number of loan purposes −4005∗∗∗ −3055∗∗∗ −3013∗∗∗ −20830 −30484 −40071 −30331 −20259 −10085∗∗ −5073∗∗∗

4101045 4101035 4100975 4205225 4201595 4401545 4200725 4207935 4504655 4104525Rating group 2 (B-letter ratings) 43026∗∗∗ 44002∗∗∗ 70752 49028∗∗∗ 10410 56068∗∗∗

4140105 4120155 4800735 4302835 4601575 4702275Rating group 3 (C-letter ratings) 13700∗∗∗ 13708∗∗∗ 56021∗∗∗ 13602∗∗∗ 91047∗∗∗ 14604∗∗∗

4180505 4260905 4180945 4601295 4210525 4900685Rating group 4 (unrated) 51055∗∗∗ 52030∗∗∗ 13071∗ 59012∗∗∗ 80823∗

4150345 4120345 4705915 4300175 4500505Number of banks −0066∗∗∗

4001665Amount −0001∗∗∗

400002605Maturity 20303∗∗∗

4006315Secured or guaranteed −40518

4308955Sales/100 −000008∗∗∗

40000000015Financial leverage 20792∗∗∗

4003835% foreign sales −000208

40003035Net income over assets −00526

4003885Property, plant, and equipment/assets −70058

4504235Observations 86,701 86,701 86,701 77,771 86,354 18,607 86,772 26,544 6,108 24,530R2 00100 00106 00111 00118 00106 00183 00186 00309 00170 00170

Notes. The dependent variable is the spread. All regressions include 21 primary loan purpose dummies, 46 loan instrument dummies, 69 currency dummies, 11 borrower typedummies, 56 borrower business (industry) dummies, year dummies, borrower nationality dummies, lead bank nationality dummies and a constant term. Regression 3 includes14 rating group dummies whose coefficients are not reported. Parameters are estimated by ordinary least squares. Standard errors are presented in parentheses and are correctedfor heteroskedasticity and clustered at the borrower nationality times lead bank nationality level.

∗∗∗, ∗∗, and ∗ denote statistical significance at the 1%, 5%, and 10% levels, respectively.

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Giannetti and Yafeh: Do Cultural Differences Between Contracting Parties Matter?376 Management Science 58(2), pp. 365–383, © 2012 INFORMS

be headquartered in the United States and the UnitedKingdom. The robustness of the results to the exclu-sion of U.S. and UK lenders, as well as to the exclu-sion of U.S. borrowers (column (8)) suggests that theeffect of cultural distance is not restricted to the inter-action between “Anglo-Saxon” economic agents andthe rest of the world. The estimated effect of culturaldistance on the spread remains unchanged even aftercontrolling for firm size (sales), financial leverage,percentage of foreign sales, profitability, and the pro-portion of tangible assets (property, plants, and equip-ment) in column (9). Thus, any remaining unobservedheterogeneity biasing our results should be uncorre-lated with any of these factors, which is unlikely.

We also run the regressions for groups of borrowerswith the same ratings, for rated borrowers (for whichinformation problems are presumably less severe) andfor loans issued in different continents. The estimates(reported only for the subsample of rated borrowersin column (10)) show that the effect of cultural dis-tance is once again unchanged. Finally, we considerwhether the effect of cultural distance changes overtime. The results are qualitatively unchanged if wedrop the loans issued during the 1980s; however, dur-ing the 1980s, the effect of cultural distance is largerthan the one we report in Table 3.

Some insights can be gained from the coeffi-cients of the control variables. It is comforting thatloan spreads are higher for borrowers with ratingsbelow A; unrated borrowers obtain credit at lowerinterest rates than borrowers with C or lower rat-ings. Furthermore, stronger creditor rights in the bor-rower’s country tend to decrease the loan cost, eventhough—unsurprisingly given that we include bor-rower nationality fixed effects—the coefficient is notstatistically significant in most specifications.

We further address the issue of unobserved het-erogeneity by using borrower fixed effects. Column(1) of Table 4 shows that, not only does the effectof cultural distance on the loan spread continue tobe positive and significant, but the magnitude of thecoefficient is similar to the one in Table 3. We alsoestimate a two-stage Heckman selection model. Thefirst-stage estimates in column (2) of Table 4 confirmthat our instruments are statistically significant: Theprobability that a loan is obtained from a given bankis decreasing in the number of physically and cultur-ally close banks. In addition, cultural distance doesnot affect the probability of a bank-borrower match,in line with empirical evidence showing that agentsunderstate the effect of cultural differences on eco-nomic outcomes (Weber and Camerer 2003). In thesecond stage, the coefficient of the inverse Mills ratiois not statistically significant, further indicating thatselection problems are not driving our results. Most

strikingly, the effect of cultural distance is now almosttwice as large as in Table 3.10

4.3. Nonprice Contract TermsAn effect of cultural differences on contracting costsshould be reflected also in more restrictive nonpriceloan terms. Estimates in Table 5 show that cultur-ally distant banks provide smaller loans (column (1))and are more likely to request loan guarantees froma third party (column (4)). These effects are also eco-nomically significant: A lender whose cultural dis-tance from the borrower is about one (roughly thecultural distance between Germany and the UnitedKingdom) is likely to receive a loan that is nearly fourmillion dollars (6.7% of the sample median) smallerthan a similar domestic borrower; the probability thata third party guarantee for the loan would be requiredis higher by about two percentage points, a largenumber given that only about 7% of the loans in thesample are guaranteed. Cultural distance also has apositive impact on the probability that the loan issecured, although the effect is not statistically signif-icant at conventional levels. We find no effect of cul-tural distance on loan maturity.

4.4. Ex Post PerformanceIf culturally distant banks had a rational concernabout attracting clients with poor credit prospects,then the loans to these borrowers should exhibit poorperformance relative to the average loan. To evalu-ate the ex post performance of borrowers, we firstconsider the probability of default. As in previous lit-erature (Altman and Suggit 2000, Emery and Cantor2005), we identify defaulting borrowers as borrow-ers that are rated when the loan is extended and arethen downgraded to a default rating. As mentionedabove, we first verify that cultural distance increasesthe cost of the loans also in this subsample. Then,we test whether the probability of default after theloan is granted is higher for borrowers receiving loansfrom culturally distant banks. Column (1) of Table 6shows that cultural distance is unrelated to the defaultprobability.

Because default rates in the syndicated loan marketare quite low, they may not fully capture the expo-sure of the lender to credit risk. We therefore explorechanges in the credit rating of borrowers after theloan is granted. A borrower’s upgrade (downgrade)

10 The standard errors we report in Table 4 are not corrected forclustering. This is because, with our large set of controls in thefixed effect and the Heckman models, Stata is unable to computeclustered standard errors. This inconvenience disappears if weexclude some controls, such as the instrument type dummies; theestimates of our variable of interest remain highly statistically sig-nificant with clustering at the borrower nationality times lead banknationality level.

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Giannetti and Yafeh: Do Cultural Differences Between Contracting Parties Matter?Management Science 58(2), pp. 365–383, © 2012 INFORMS 377

Table 4 Addressing Selection Problems

(1) (2) (3)

Borrower fixed effects Heckman selection model

Spread Bank-firm match Spread

Cultural distance 11007∗∗ −000178 17059∗∗

4501295 4000125 470035Distance −00247 00000702 30275∗∗

4008555 40000265 410435Same religion −60475 000164 −35091∗∗∗

4602725 4000185 490475Bank rank −00102∗∗∗

4000155Number close banks −00790∗∗∗

4000325Border×Number of close banks −000791∗∗∗

40000615Number of culturally distant banks 00109∗∗∗

40000735Border×Number of culturally distant banks −000859∗∗∗

40000675Mills ratio −170786

41402985Observations 86,354 350,411 15,963Wald chi-squared 7,726.77R2 00502

Notes. Column (1) presents estimates obtained by controlling for borrower fixed effects. Columns (2) and (3) report theestimates of a Heckman selection model. In column (2) (first stage), we consider how a borrower is matched to all top500 potential lead banks in the country; the unit of analysis is the potential borrower-lead bank match and the dependentvariable is a dummy that takes the value one if a borrower receives a loan from a given lead bank that has been operationalin its country in the past, and equals zero if the borrower does not receive a loan from that lead bank. In column (3), weconsider all loans issued by the top 500 lead banks in country. In addition to variables defined in Table 1, the selectionequation in column (2) includes the rank of the lead bank in a country according to the number of deals concluded up tothe year of the loan, the number of close foreign banks (foreign banks from countries with a capital city less than 2,000km from the capital city of the country of the borrower), and the number of culturally distant foreign banks (foreign bankswith cultural distance above the median cultural distance from the country of the borrower). In addition to the reportedcoefficients, we include all the control variables used in column (5) of Table 3 as well as 21 primary loan purpose dummies,46 loan instrument type dummies, 69 currency dummies, year dummies, dummies, lead bank nationality dummies, and aconstant term. Standard errors are presented in parentheses.

∗∗∗, ∗∗, and ∗ denote statistical significance at the 1%, 5%, and 10% levels, respectively.

indicates that its credit quality has improved (dete-riorated) after the extension of the loan and priorto its maturity.11 In column (2), we present estimatesof an ordered probit model in which we considerobtaining a rating as an upgrade and losing a rat-ing as a downgrade. Strikingly, after controlling forloan and borrower country characteristics, culturallydistant borrowers are more likely to be upgraded,not downgraded. This confirms that the loan termsoffered by culturally distant banks are not justified bythe borrowers’ poor credit prospects.

The estimates in column (2) include unrated firmsand cases where such firms obtain a rating (lose a

11 An upgrade cannot be interpreted as incorporating positiveinformation generated by the granting of the loan because thisinformation is already incorporated in the borrower’s rating whenthe loan is granted.

rating) are treated as upgrades (downgrades). In col-umn (3), we restrict the sample to firms with actualrating changes. The estimates again suggest no sys-tematic effect of cultural distance. This result holdsalso when a rating change refers not only to lettergrade changes but also to changes in notches withinthe same letter grade.12

12 We also estimate separate probit models for upgrades and down-grades. For upgrades (downgrades) we define the dependent vari-able to be equal to one if the borrower is upgraded (downgraded)by at least a notch and equal to zero if the borrower’s rating isunchanged or if the borrower is downgraded (upgraded). Includ-ing the same controls as in the ordered probit models in Table 6,we find that a marginal increase in cultural distance increases theprobability of an upgrade by 16% (the effect is statistically signif-icant at 5%). By contrast, an increase in cultural distance has nostatistically significant effect on the probability of a downgrade.

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Giannetti and Yafeh: Do Cultural Differences Between Contracting Parties Matter?378 Management Science 58(2), pp. 365–383, © 2012 INFORMS

Table 5 Determinants of Other Contractual Features

(1) (2) (3) (4)Amount Maturity Secured Guaranteed

Cultural distance −30877∗∗ 000482 0000517 000202∗∗

4105515 40007945 400007285 400008755Distance 00453∗ −000177 −0000130 00000205

4002455 40001665 4000009535 400001225Same religion 30159∗ −00215∗∗ −000113 −000122

4107555 4001085 400009235 400008095Observations 116,803 101,202 117,194 117,194Adjusted R2 00312 00380 00250 00200

Notes. The dependent variables are loan amount, loan maturity, and binaryvariables denoting secured or guaranteed loans. In column (1), where theloan amount appears as a dependent variable, the largest observations arewinsorized at the 1% level. In addition to the reported coefficients, we includeall the control variables used in column (5) of Table 3 as well as 21 primaryloan purpose dummies, 46 loan instrument dummies, 69 currency dummies,11 borrower type dummies, 56 borrower business dummies, year dummies,borrower nationality dummies, lead bank nationality dummies, and the con-stant term (coefficients not reported for brevity). Parameters are estimatedby ordinary least squares. Standard errors are presented in parentheses andare corrected for heteroskedasticity and clustered at the borrower nationalitytimes lead bank nationality level.

∗∗∗, ∗∗, and ∗ denote statistical significance at the 1%, 5%, and 10% levels,respectively.

Finally, because ratings and their changes are noisyproxies for borrower performance, we turn to thesubsample of loans with firm-level data from World-scope, in which, as shown in Table 3, we find a pos-itive effect of cultural distance on the loan spread.

Table 6 Ex Post Borrower Performance

(1) (2) (3) (4) (5) (6) (7) (8)

Default Changes in rating Changes in borrower characteristics

Rated Acquiring Actual Actualcompanies rating = rating rating changes; Market to

only Upgrade changes finer ratings book Leverage ROA Sales

Cultural distance −00000501 000284∗ 0000362 −000241 00235 000254 00129 000192400001875 40001545 40002595 40002435 4003945 40007345 4002315 4001985

Observations 9,943 41,336 9,943 9,943 1,970 2,093 1,989 2,090R2 0019 0030 0026 0006 00034 00040 00026 00095

Notes. The dependent variables are measures of borrower performance after the loan is granted. In column (1), the dependent variableis a dummy that takes value 1 if a borrower’s rating is changed to default before the maturity of the loan and equal to zero if theborrower continues to have a no default rating; parameter estimates are obtained using a probit model. In column (2), the dependentvariable takes the value 1 4−15 if the borrower was upgraded (downgraded) by Moody’s or S&P after the loan issuance and beforeits maturity and the value zero if the rating remained unchanged; obtaining (losing) a rating is treated as an upgrade (downgrade). Incolumn (3), the dependent variable takes the value 1 (−1) if the borrower was upgraded (downgraded) by Moody’s or S&P after theloan issuance and before its maturity and the value zero if the rating remained unchanged; we exclude unrated borrowers. Whereas incolumns (2) and (3), we consider as a change in rating only changes in letter grades, in column (4), we consider changes in notches;we exclude unrated borrowers. In columns (2)–(4), estimates are obtained using an ordered probit model. In columns (5)–(8), thedependent variable is the change in the measure of borrower performance indicated on top of the column during the two years afterthe loan is granted. Parameter estimates in these columns are obtained by ordinary least squares. In all equations, we include thefollowing control variables whose coefficients are not reported: year dummies, borrower type dummies, the borrower’s initial ratinggroup, GDP in the country of the borrower, and the time since the loan was issued. Standard errors are presented in parentheses andare corrected for heteroskedasticity and clustered at the borrower nationality times lead bank nationality level.

∗∗∗, ∗∗, and ∗ denote statistical significance at the 1%, 5%, and 10% levels, respectively.

Table 6 shows that, for these borrowers, cultural dis-tance is unrelated to changes in the market to bookratio, leverage, sales, and profitability in the two yearsfollowing the issuance of the loan (in unreportedspecifications, we show that this is the case also oneyear, three years, and four years after the issuance ofthe loan). Because changes in creditworthiness shouldbe related to changes in firm value or accounting per-formance, this strongly suggests that clients of cultur-ally distant banks are as creditworthy as the clients ofother banks.

4.5. Local Subsidiaries and Repeated InteractionWe now explore how the effect of cultural distancevaries with the lender’s experience in the borrower’scountry or with repeated borrower–lender interac-tions. Banks with a subsidiary in the borrower’scountry should have more within-country experiencebecause they tend to have extended a larger numberof loans in that country and have at least some localemployees. The effect of culture may nevertheless per-sist if the managers of the subsidiary in charge ofapproving the loans are from the headquarters’ coun-try or if the culture of the country of origin affects thesubsidiary’s organization.

Table 7 shows that having a local subsidiary inthe country of the borrower mitigates, but does noteliminate the effect of cultural distance. The effectsof cultural distance on the spread (column (1)) andthe probability of having a loan guarantor are almost

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Giannetti and Yafeh: Do Cultural Differences Between Contracting Parties Matter?Management Science 58(2), pp. 365–383, © 2012 INFORMS 379

Table 7 Local Subsidiaries

(1) (2) (3) (4) (5)Spread Amount Maturity Secured Guaranteed

Cultural distance 10009∗∗ −2038 0012 −000036 00023∗∗∗

450015 4106375 400085 40000785 4000095Cultural distance× Local subsidiary −4013∗ −4064∗∗ −0021∗∗∗ 00027∗∗∗ −00009∗

420135 410965 400065 4000095 4000055Distance 00005 0050∗∗ −000155 −000016 000003

4008235 400255 40001625 400000975 40000125Same religion −7037 3071∗∗ −0019∗ −00015 −00011

440695 410885 400115 4000105 4000085Observations 86,354 116,803 101,202 117,194 117,194Adjusted R2 0011 0031 00380 00250 00200

Notes. The dependent variables are spread, amount, maturity, and binary variables denoting secured or guaranteedloans. In addition to the reported coefficients, we include all the control variables used in column (5) of Table 3 aswell as 21 primary loan purpose dummies, 46 loan instrument type dummies, 69 currency dummies, 11 borrowertype dummies, 56 borrower business dummies, year dummies, borrower nationality dummies, lead bank nationalitydummies, and a constant term (coefficients not reported for brevity). In addition to the previously defined variables,Local subsidiary is a dummy variable that takes the value one if the lead bank has a local subsidiary in the countryof the borrower and zero otherwise. Parameters are estimated by ordinary least squares. Standard errors arepresented in parentheses and are corrected for heteroskedasticity and clustered at the borrower nationality timeslead bank nationality level.

∗∗∗, ∗∗, and ∗ denote statistical significance at the 1%, 5%, and 10% levels, respectively.

Table 8 The Dynamics of Cultural Biases

(1) (2) (3) (4) (5)Spread Amount Maturity Secured Guaranteed

Cultural distance 11045∗∗∗ −20723 −000127 0000978 000175∗

4402335 4107725 40009195 400008145 40001045Borrower interaction 00919 −10503∗∗∗ −000735∗∗∗ 0000432∗∗∗ −0000232∗

4008555 4001545 400007285 4000008165 400001215Cultural distance×Borrower interaction −20999∗∗∗ −00802∗ 000425∗ −00000239 0000125

4009455 4004735 40002445 400002245 400003935Distance −00182 00545∗∗ −000232 −0000198∗ 00000290

4006015 4002725 40002055 400001125 400001415Same religion 00182 20046 −00295∗∗ −000109 −0000809

4406935 4200045 4001275 40001075 400008775Observations 79,022 105,433 91,892 105,753 105,753Adjusted R2 00102 00321 00375 00227 00183

Notes. The dependent variables are spread, amount, maturity, and binary variables denoting secured or guaranteedloans. We consider only syndicated loans made starting in 1990. In addition to the reported coefficients, we includeall the control variables used in column (5) of Table 3 as well as 21 primary loan purpose dummies, 46 loan instru-ment dummies, 69 currency dummies, 11 borrower type dummies, 56 borrower business dummies, year dummies,borrower nationality dummies, lead bank nationality dummies, and a constant term (coefficients not reported forbrevity). Parameters are estimated by ordinary least squares. Standard errors are presented in parentheses and arecorrected for heteroskedasticity and clustered at the borrower nationality times lead bank nationality level.

∗∗∗, ∗∗, and ∗ denote statistical significance at the 1%, 5%, and 10% levels, respectively.

halved (column (5)). The negative effect of culturaldistance on the size of the loan is, however, magni-fied. Furthermore, culturally distant banks with localsubsidiaries grant loans with shorter maturity and aremore likely to secure the loan. Because short matu-rity and collateral are useful if the lender monitors theborrower, this finding suggests that the lender’s expe-rience makes monitoring less costly. The reduction inthe marginal cost of monitoring appears to dominate

any decrease in contracting costs, which should haveled to less restrictive contract terms.13

In Table 8, we find some evidence that repeatedinteraction with a given borrower mitigates the effectof cultural distance. To avoid biases deriving from

13 We also examine the effect of lead bank experience in the bor-rower’s country and find little evidence that the effect of culturaldistance disappears after the lead bank has concluded many dealsthere (results not tabulated).

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Giannetti and Yafeh: Do Cultural Differences Between Contracting Parties Matter?380 Management Science 58(2), pp. 365–383, © 2012 INFORMS

the fact that previous interactions are, by construc-tion, very few at the beginning of the sample period,we only include loans signed on or after 1990. Forthe effect of cultural distance on the spread to dis-appear, the borrower has to receive nearly four syn-dicated loans from a given lead bank. However, in95% of the loans in the sample, the borrower receivedat most two previous loans from a given lead bank(the median number of loans from a given bankis one).14 Thus, the effect of cultural differences isonly partially mitigated by repeated borrower–lenderinteraction. Repeated interaction with a culturally dis-tant lead bank appears to enable the borrower toreceive loans with longer maturity; however, it doesnot increase the size of the loan and has no signifi-cant impact on the probabilities that collateral or thirdparty guarantees are required. Given that its effectpersists despite repeated interaction, cultural distanceis unlikely to exclusively capture information gather-ing costs, which should be most relevant for first-timeborrowers.

A possible concern is that the estimates in Table 8are driven by unobserved risk profiles characteriz-ing repeated borrowers from culturally distant banks.To evaluate the merit of this alternative explana-tion, in unreported specifications, we test whetherthe effect of cultural distance on the probability ofloan default differs between first-time and repeatedborrowers: Cultural distance continues to have aninsignificant effect on the probability of default inboth subsamples; this confirms once again that theeffect of cultural distance does not depend on bor-rower unobserved heterogeneity.

4.6. Further RobustnessAlthough our results so far indicate that the effect ofcultural distance is unlikely to be driven by borrowerheterogeneity, concerns may remain that cultural dis-tance is related to some characteristics of the countrypair. For instance, low levels of international trade orinvestment may be correlated with cultural distanceand lead to limited information flows between cer-tain countries. However, our results hold even whenwe control for bilateral trade or investment flows orfor industrial similarity of the lender–borrower coun-try pair (see Giannetti and Yafeh 2010). The resultsalso hold if we exploit only the limited time-seriesvariation of cultural distance by including borrowernationality times lender nationality fixed effects.

14 Interestingly, in unreported regressions we find that the numberof loans that a borrower receives from a given lead bank decreaseswith cultural distance. Our main results, however, are not drivenby the fact that borrowers are less likely to engage culturally distantbanks in repeated relations: The effect of cultural distance on theloan spread is larger if we consider only the first loan a borrowerreceives from any given bank.

Finally, we try to shed some further light on themechanism through which cultural differences affectfinancial contracts. If culturally distant banks are lessinformed than other banks, under the hypothesis,strongly supported by the empirical evidence, thatthe clients of culturally distant banks are similar tothe clients of other banks, the variance of the con-tract terms that culturally distant banks offer to bor-rowers in a given country should be lower than thevariance of contract terms of culturally close banks.In unreported specifications, we find no evidence ofthat, suggesting that culturally distant banks are asinformed as other banks.

5. Cultural Distance Between Banksand Risk Sharing Within theSyndicate

If cultural differences affect interactions betweeneconomic agents, we should observe their effectsalso on the interaction between lead banks andparticipant banks. Keeping culture constant, nego-tiations may be faster for a smaller investment. Ifculture increases negotiation costs, participant banksmay buy a smaller share of the loan to reduce thenegotiation time. Ceteris paribus, a negative effect onthe difference between the share of the loan bought bya participant bank and the share of the loan retainedby the lead bank would suggest that cultural dis-tance negatively affects negotiations and reduces risksharing.

Because our unit of analysis is the extent of risksharing between the lead bank and each participantbank, and given that each loan has, on average,several participant banks, we have multiple obser-vations for each loan. For this reason, we clusterstandard errors at the loan level.15 The results showthat, indeed, participant banks hold smaller portionsof loans syndicated by culturally remote lead banks.In column (1) of Table 9, a one-standard-deviationincrease in cultural distance, approximately the dif-ference between United States and Canada, decreasesrisk sharing between two banks by nearly 5% (rel-ative to the sample mean). The effect is even morepronounced if we exclude observations for which thelead and participant banks share the same national-ity (column (2)). In this case, a one-standard-deviationincrease in the cultural distance decreases risk shar-ing by over 10%. These results are consistent with thenotion that cultural differences increase contractingcosts, but harder to explain with an omitted factor:There is no reason to believe that an omitted factor

15 The statistical significance of the results is similar if we clustererrors at the lead bank nationality times participant bank national-ity level.

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Giannetti and Yafeh: Do Cultural Differences Between Contracting Parties Matter?Management Science 58(2), pp. 365–383, © 2012 INFORMS 381

Table 9 Risk Sharing Within the Syndicate

(1) (2) (3) (4) (5) (6) (7) (8)Loan FE; Excluding Excluding

Whole Foreign whole U.S. U.S. lead Foreign Foreign Foreignsample participants sample borrowers banks participants participants participants

Banks’ cultural distance −00162∗∗∗ −00334∗∗∗ −000367∗∗ −00408∗∗∗ −00377∗∗∗ −00300∗ −00493∗∗∗ −00387∗

4000565 400125 4000155 4000825 4000795 400175 400145 400205Banks’ cultural distance 000138∗∗ 000129∗

× Interaction syndicate 40000685 40000695Interaction syndicate −000222∗ −000237∗

4000125 4000135Banks’ distance −00122∗∗∗ −00132∗∗∗

4000265 4000295Same religion syndicate 00456∗∗∗ 00450∗∗

400175 400195Observations 225,704 114,159 227,752 115,522 124,073 114,049 101,656 101,562Adjusted R2 0003 0003 0089 0007 0007 0003 0003 0003

Notes. The dependent variable is risk sharing. For each loan we have a number of observations equal to the number of participant banks. In columns (2) and(6)–(8), we include only observations for which the nationality of the lead bank is different from the nationality of the participant bank (foreign participants).Additionally, in the regressions in which we include the number of bank interactions (columns (7) and (8)), we consider only syndicated loans made since1990. All regressions include controls for per-capita GDP in the lead and participant banks’ countries, controls for the loan rating, for whether the loan istranched, for the number of loan purposes as well as 21 primary loan purpose dummies, 46 loan instrument type dummies, 69 currency dummies, 11 borrowertype dummies, 56 borrower business dummies, year dummies, borrower nationality dummies, lead bank nationality dummies, participant bank nationalitydummies and a constant term. In addition, in columns (6) and (8) we control also for common border, legal origin, language, and colonial ties in the lead andthe participant banks’ countries, creditor rights in both countries, and the absolute value of their difference (coefficients not reported for brevity). Parametersare estimated by ordinary least squares. Standard errors are presented in parentheses and are corrected for heteroskedasticity and clustered at the loan level.

∗∗∗, ∗∗, and ∗ denote statistical significance at the 1%, 5%, and 10% levels, respectively.

should be similarly correlated with the cultural dis-tance between borrowers and lenders and with thecultural distance between lead banks and participantbanks.

Because we have multiple observations for eachloan, we can perform a more stringent test for unob-served borrower heterogeneity. In column (3), weinclude loan fixed effects. The estimates show that,even for the same loan, culturally distant participantsshare less risk with the lead bank than culturallycloser participants. The effect of cultural distance isrobust across different samples. For instance, in col-umn (4), we exclude loans to U.S. borrowers, andin column (5), we exclude loans for which the leadbank is from the United States. Similarly, the coeffi-cient of cultural distance is qualitatively unchangedin column (6), when we include additional controlsfor distance and investor protection.

Some of the control variables offer further interest-ing insights. Risk sharing is higher if the participantbank is from a country with the same religion as thatof the lead bank’s country, but is significantly lowerif banks are from physically remote countries: A one-thousand kilometer increase in distance decreases risksharing by 12 percentage points.16

16 In unreported specifications, we also control for the size of thelead bank and the participant bank in terms of the syndicated loansthey held during the previous year. As expected, large lead banks

In columns (7) and (8), we explore whether theeffect of cultural distance on risk sharing declines asa bank participates in more deals with a given leadbank. We focus on interactions within a country tocapture the possibility that employees responsible fora given country may learn to interact with the repre-sentatives of the lead bank in that country. In this casetoo, in order to avoid biases resulting from the factthat previous interactions are, by construction, veryfew at the beginning of the sample period, we onlyinclude loans signed on or after 1990. We find that,indeed, the effect of cultural distance becomes smalleras the number of previously concluded deals with agiven lead bank increases. Nevertheless, the pace atwhich the negative effect of cultural differences diesout is very slow, and over 30 deals are needed to fullyoffset the effect of cultural distance on risk sharing.The mean (median) number of deals that a participantconcludes with a given lead bank is, however, onlyeight (two).

6. ConclusionThis paper shows that professional decision makersare inclined to offer better terms to culturally sim-ilar counterparties. Not only do cultural differences

share risk less; however, the size of the participant bank does notseem to affect its portion of the loan. More importantly, the effectof cultural distance on risk sharing is unchanged.

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Giannetti and Yafeh: Do Cultural Differences Between Contracting Parties Matter?382 Management Science 58(2), pp. 365–383, © 2012 INFORMS

between borrower and lender increase the loan spreadand limit the loan size, cultural differences betweenthe lead bank and participant banks reduce the extentof risk sharing within the syndicate. Although wecannot provide a definitive statistical proof that self-selection problems and unobserved heterogeneity arenot driving our findings, we perform a battery oftests consistently showing no evidence that culturaldistance is related to loan and borrower characteris-tics that negatively affect loan terms and risk sharing.We therefore propose that cultural differences makenegotiations more cumbersome and increase contract-ing costs. Future work exploring the relevance ofthe effects we document to other contexts, perhapsthrough controlled experiments, will be able to castmore light on the effects of cultural differences oncontractual outcomes.

Our findings also suggest new avenues for boththeoretical and empirical research in behavioral eco-nomics. Although we document that cultural dif-ferences adversely affect contract terms, we areunable to describe the precise mechanisms throughwhich differences in codes and norms affect contrac-tual outcomes. Theories and hypotheses on potentialmechanisms would be useful in guiding future empir-ical research with the goal of identifying the mecha-nisms at work.

AcknowledgmentsThe authors thank Ronald Inglehart for sharing some ofthe WVS data. For comments, the authors thank BradBarber (special issue coeditor); two anonymous referees;Daniel Baraz; Sreedhar Bharath; Mike Burkart; RadhaGopalan; Tong Hui; Eugene Kandel; Amir Licht; EricNowak; Enrichetta Ravina; Lilach Sagiv; and seminar par-ticipants at the American Finance Association Conferencein Atlanta, the Frontiers of Corporate Finance and Gover-nance Conference in Banff, the Washington University in St.Louis Corporate Finance Conference, the European FinanceAssociation Conference in Frankfurt, the Indian School ofBusiness Summer Research Conference, Michigan State Uni-versity, Southern Methodist University, INSEAD, NanyangTechnological University, the National University of Singa-pore, the Stockholm School of Economics, the InternationalMonetary Fund, the University of Naples “Federico II,”the Einaudi Institute for Economic and Finance (EIEF), BarIlan University, the Hebrew University, the Leuven-LouvainFinance Workshop, and the University of Mannheim. Theauthors are also grateful to Daniella Gera, Max Gladstein,and Shai Harel for excellent research assistance. Marias-sunta Giannetti acknowledges financial support from theSwedish National Research Council and the Jan Wallanderand Tom Hedelius Foundation. Part of the work on thisproject was carried out while Yishay Yafeh was residentscholar at the IMF Research Department, whose hospital-ity is gratefully acknowledged. Yishay Yafeh also acknowl-edges financial support from the Krueger Center at theHebrew University.

ReferencesAdair, W., J. Brett, T. Okumura. 2001. Negotiation behavior when

cultures collide: The United States and Japan. J. Appl. Psych.86(3) 371–385.

Ahern, K., D. Daminelli, C. Fracassi. 2010. Lost in translation? Theeffect of cultural values on mergers around the world. Mimeo,University of Michigan, Ann Arbor.

Alesina, A., F. Lotti, P. E. Mistrulli. 2008. Do women pay morefor credit? Evidence from Italy. NBER Working Paper 14202,National Bureau of Economic Research, Cambridge, MA.

Altman, E., H. Suggitt. 2000. Default rates in the syndicated bankloan market: A mortality analysis. J. Banking Finance 24(1–2)229–253.

Bae, K., V. Goyal. 2009. Creditor rights, enforcement and bankloans. J. Finance 64(2) 823–860.

Becker, G. 1971. The Economics of Discrimination. University ofChicago Press, Chicago.

Bloom, N., R. Sadun, J. Van Reenen. 2009. The organization of firmsacross countries. NBER Working Paper 15129, National Bureauof Economic Research, Cambridge, MA.

Bottazzi, L., M. Da Rin, T. Hellmann. 2007. The importance of trustfor investment: Evidence from venture capital. Mimeo, Univer-sity of British Columbia, Vancouver.

Bottazzi, L., M. Da Rin, T. Hellmann. 2008. Who are the activeinvestors? Evidence from venture capital. J. Financial Econom.89(3) 488–512.

Brett, J., T. Okumura. 1998. Inter- and intracultural negotiation: U.S.and Japanese negotiators. Acad. Management J. 41(5) 495–510.

Carey, M., G. Nini. 2007. Is the corporate loan market globally inte-grated? A pricing puzzle. J. Finance. 62(6) 2969–3007.

Coval, J., T. Moskowitz. 1999. Home bias at home: Local equitypreference in domestic portfolios. J. Finance 54(6) 2045–2073.

Dennis, S., D. Mullineaux. 2000. Syndicated loans. J. Financial Inter-mediation 9(4) 404–426.

Djankov, S., C. McLiesh, A. Shleifer. 2007. Private credit in 129 coun-tries. J. Financial Econom. 84(2) 299–329.

Duffie, D., N. Garleanu, L. Pedersen. 2005. Over-the-counter mar-kets. Econometrica 73(6) 1815–1847.

Duffie, D., N. Garleanu, L. Pedersen. 2007. Valuation in over-the-counter markets. Rev. Financial Stud. 20(6) 1865–1900.

Emery, K., R. Cantor. 2005. Relative default rates on corporate loansand bonds. J. Banking Finance 29(6) 1575–1584.

Esty, B., W. Megginson. 2003. Creditor rights, enforcement, and debtownership structure: Evidence from the global syndicated loanmarket. J. Financial Quant. Anal. 38(1) 37–59.

Giannetti, M., Y. Yafeh. 2010. Do cultural differences betweencontracting parties matter? Evidence from syndicatedbank loans. Working paper, Stockholm School of Eco-nomics, Stockholm. http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1204202.

Grinblatt, M., M. Keloharju. 2001. How distance, language, andculture influence stockholdings and trades. J. Finance 56(3)1053–1073.

Guiso, L., P. Sapienza, L. Zingales. 2006. Does culture affect eco-nomic outcomes. J. Econom. Perspect. 20(2) 23–48.

Guiso, L., P. Sapienza, L. Zingales. 2009. Cultural biases in economicexchange. Quart. J. Econom. 124(3) 1095–1131.

Hofstede, G. 2001. Culture’s Consequences: Comparing Values, Behav-iors, Institutions, and Organizations Across Nations, 2nd ed. SagePublications, Thousand Oaks, CA.

Huberman, G. 2001. Familiarity breeds investment. Rev. FinancialStud. 14(3) 659–680.

Inglehart, R. 1997. Modernization and Postmodernization: Cultural,Economic, and Political Change in 43 Societies. Princeton Univer-sity Press, Princeton, NJ.

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Giannetti and Yafeh: Do Cultural Differences Between Contracting Parties Matter?Management Science 58(2), pp. 365–383, © 2012 INFORMS 383

Inglehart, R., W. Baker. 2000. Modernization, cultural change, andthe persistence of traditional values. Amer. Sociol. Rev. 65(1)19–51.

Inglehart, R., C. Welzel. 2005. Modernization, Cultural Change andDemocracy. Cambridge University Press, New York.

Kisgen, D. 2006. Credit ratings and capital structure. J. Finance 61(3)1035–1072.

Kogut, B., H. Singh. 1988. The effects of national culture on thechoice of entry mode. J. Internat. Bus. Stud. 19(3) 411–432.

Mead, R., T. Andrews. 2009. International Management. John Wiley& Sons, West Sussex, UK.

Qian, J., P. Strahan. 2007. How law and institutions shape financialcontracts: The case of bank loans. J. Finance 62(6) 2803–2834.

Ravina, E. 2008. Love & Loans. The effect of beauty and personalcharacteristics in credit markets. Mimeo, Columbia University,New York.

Rogers, E., D. Bhowmik. 1970. Homophily-heterophily: Relationalconcepts for communication research. Public Opinion Quart.34(4) 523–538.

Rose, A. 2004. Do we really know that the WTO increases trade?Amer. Econom. Rev. 94(1) 98–114.

Schwartz, S. 2006. A theory of cultural value orientations: Explica-tion and applications. Comparative Sociol. 5(2–3) 137–182.

Siegel, J. I., A. N. Licht, S. H. Schwartz. 2011. Egalitarianism andinternational investment. J. Financial Econom. 102(3) 621–642.

Sorensen, M. 2007. How smart is smart money? A two-sided match-ing model of venture capital. J. Finance 62(6) 2725–2762.

Weber, R. A., C. F. Camerer. 2003. Cultural conflict and merger fail-ure: An experimental approach. Management Sci. 49(4) 400–415.

Weber, Y., O. Shenkar, A. Raveh. 1996. National and corporate cul-tural fit in merger/acquisitions: An exploratory study. Manage-ment Sci. 42(8) 1215–1227.

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.